Mixed methods analysis of enterprise social networks

Computer Networks xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Computer Networks journal homepage: www.elsevier.com/locate/comnet

Mixed methods analysis of enterprise social networks Sebastian Behrendt a,⇑, Alexander Richter b, Matthias Trier c a b c

Bundeswehr University, Munich, 85577 Neubiberg, Germany University of Zurich, CH-8050 Zürich, Switzerland Copenhagen Business School, DK-2000 Frederiksberg, Denmark

a r t i c l e

i n f o

Article history: Received 19 December 2013 Received in revised form 11 July 2014 Accepted 11 August 2014 Available online xxxx Keywords: Mixed methods Social software Enterprise social networks Evaluation Social network analysis Case study

a b s t r a c t The increasing use of enterprise social networks (ESN) generates vast amounts of data, giving researchers and managerial decision makers unprecedented opportunities for analysis. However, more transparency about the available data dimensions and how these can be combined is needed to yield accurate insights into the multi-facetted phenomenon of ESN use. In order to address this issue, we first conducted a systematic literature review to identify available data dimensions and integrated them into a conceptual framework. We then adopted this framework as part of a mixed methods research approach to comprehensively analyze an empirical ESN case. With our results serving as a proof of concept we show the insights that can be derived from different data dimensions and how combining these can improve the validity of the analysis. The application of the framework also allows us to derive a detailed guideline for combining different data sources in ESN analysis to support researchers and decision makers. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction Over the past few years, many companies have witnessed their employees moving toward more flexibility and self-determination, especially in terms of spatial and time-independent work [1,2]. Systems that support distributed and networked collaboration, such as enterprise social networks (ESN), have been playing an increasingly important role in changing communication practices [3– 5]. Based on previous research [6–9], we define ESN as web-based intranet platforms that support users in contributing persistent objects to a shared pool, which enables public responses to these objects, allows profile information to be presented, and connects users via features like Following, or Friendship request. Examples of widely adopted platforms include weblogs, wikis, microblogs, ⇑ Corresponding author. Tel.: +49 89 6004 2613. E-mail addresses: [email protected] (S. Behrendt), arichter@ifi. uzh.ch (A. Richter), [email protected] (M. Trier).

and social networking platforms, which tend to converge into an integrated ESN [10]. The increasing use of ESN produces a considerable amount of data, since almost every interaction in the system leaves a persistent digital trace [3]. This ‘‘revolution in the measurement of collective human behavior’’ [11 p. 66] gives researchers and managerial decision makers unprecedented opportunities to analyze and explain such systems. The comprehensive variety of relevant ESN data is evidenced, for example, by studies on patterns of information exchange [12,13] and their underlying network structures [14,15], the distribution of ideas [16], rumors [17] and sentiments [18], as well as their representation [19]. Other studies have shed light on social technology’s effects on employee performance [20,21], hierarchy [22], contribution behavior [23], and trust [24]. Whereas existing studies have led to insightful results, they have all applied only partial views based on different individual data dimensions and research approaches.

http://dx.doi.org/10.1016/j.comnet.2014.08.025 1389-1286/Ó 2014 Elsevier B.V. All rights reserved.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

2

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

However, the abundant data suggests that systematically combining and triangulating different data dimensions will lead to more robust or even novel insights for researchers or managers. For example, conducting an additional sentiment analysis of the postings and qualitative user interviews might support interpreting dissemination measures. In addition to the challenge of combining multiple such data sources, consistent academic theory development faces important methodological challenges. Researchers have argued that the application of current social network theories to online social networks is limited in that new and unknown aspects could affect the results of ESN analysis [10]. Trier [25], also emphasizes online social interaction’s high volatility and its resulting fast structural changes. Furthermore, the digital traces within an ESN are of a different nature than the data analyzed in earlier social network studies and may lack validity under certain aspects [26]. To address the above issues and in order to achieve consistent insights across future ESN research projects, more transparency of data dimensions and their interrelationships is required. This gap motivates our first research question (RQ): RQ1. Which data dimensions are adopted in existing studies on enterprise social networks? Contrary to the abundance of available data, analytical metrics, and perspectives on the data, there is a paucity of research that systematically develops analytical methods that effectively utilize this rich resource to inform academic inquiry and managerial decision making. In order to analyze ESN data, we explore and develop a conceptual framework as part of a dedicated mixed methods approach [27,28] to address the diversity of data and the critical validity of online data traces. Mixed methods research builds on a combination of qualitative and quantitative methods [29] and is particularly appropriate for domains marked by diversity [30]. In our context, this approach is appropriate for several reasons: ESN are a multi-dimensional and complex domain [31] that only one method is unlikely to capture appropriately. Basic social network analysis (SNA) alone is particularly insufficient due to new user behavior in online networks [5,32,10]. Furthermore, mixing methods allows researchers to draw on the strength of the methods used and to offset their weaknesses [33,34]. It also helps to

clarify the results that one method provides with the results of others [34], which expands the results and enhances the integrity of the findings [33]. To contribute to a better understanding of ESN analysis, our second research question thus focusses on a first step toward a dedicated mixed ESN research method approach: RQ2. What are the implications of combining different data dimensions in a mixed methods approach in the context of an enterprise social network analysis? By answering these questions, our contributions are threefold: We first identify relevant data dimensions based on a systematic literature review and suggest a framework to improve the transparency of the available options for analysis, which can serve as a basis for future ESN analysis. In order to not limit our contribution to conceptual work, we further apply and systematically evaluate our framework empirically, using an extensive and multidimensional case analysis of an ESN for military medical personnel as a proof of concept. To our knowledge, there is as yet no such empirical investigation of different data dimensions by means of a case study. In the context of this paper, we only reflect on the methodological aspects of the case study and its different data dimensions. Based on the empirical case analysis, we shed light on the unexploited potential, but also on the limitations of a mixed methods approach in the context of an ESN analysis. Third, to exploit the benefits and to address the limitations of this approach, we develop a set of guidelines for conducting an effective mixed methods analysis of the identified data dimensions that will be helpful to researchers and practitioners alike. Fig. 1 summarizes our research questions and contributions. As we specifically intend to combine a theory-based concept development and an empirical proof of concept, the remainder of this paper is structured as follows: The next section contains the results of a structured literature review of current research into social network analysis, focusing on the research methods and utilized data dimensions. From this review, we systematically categorize all available data dimensions. Thereafter, in Section 4, we assess our framework [36] and investigate the identified dimensions and their interplay in an empirical case study, using a dedicated mixed methods approach. The presentation of the analysis results follows in Section 5. Based on the empirical results, our approach is discussed in

Fig. 1. Research questions and contributions.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

3

Section 6; we also discuss mixed methods analysis’s observed issues, as well as present guidelines that address these issues, before we conclude this paper.

corresponding data sources. We give a brief overview of the studied articles in the following section. Based on the findings of our literature review, we then develop a framework of ESN data dimensions.

2. Identifying data dimensions in ESN research

2.1. Literature review

As noted in the introduction, the existing studies on the multi-facetted phenomenon of ESN use are based on a variety of data sources and methods of data collection. Currently, there seems to be a lack of transparency about the available options and likely implications of concrete decisions on the sources that should be adopted in a study. To address this gap, we first conducted a systematic literature review, which aimed to identify, categorize, and conceptualize the variety of research measures, analytical methods, and data sources used in recent academic work. We then developed a framework of data dimensions based on this review. Our research is focused at examining how ESN as information systems can be understood, analyzed and subsequently utilized by actors, such as researchers but also managerial decision makers. The corresponding discipline that explicitly addresses this vantage point is information systems research. The association of information systems researchers (AIS) recommends eight accepted top IS journal outlets in their AIS Senior Scholars’ Basket of Journals [37]. We used this list as a starting point for our literature review, as the respective journals primarily follow a phenomenological interest and focus on theoretical explanations, rendering them very suitable for our research objectives. We limited our search to the keywords ‘‘online social networks AND analysis’’ and ‘‘social network analysis.’’ Since the topic of online social networks has only emerged recently, we had no limitation regarding the timeframe. We thus derived a preliminary set of 37 relevant papers, of which JMIS and MISQ produced the most contributions. After reading and analyzing the abstracts, we disregarded papers not related to our core domain, that is, ESN. This resulted in 13 initial papers, published between 2010 and 2014, which we considered a starting point for a forward and backward search1 [38] to identify further important papers that could extend our coverage. Thereby, we also extended our corpus to papers outside the AIS Senior Scholars’ Basket of Journals. Furthermore, we included papers published at major conferences. Our goal was to extend the scope of our analysis and the diversity of the adopted data dimensions. After this second step, we derived a final data set of 22 relevant papers, with the earliest paper being from 2005. Based on a detailed reading, we identified the focal research measure(s), the adopted research methods, and the data sources. The publication years clearly indicate the novelty of this research domain within the AIS Senior Scholars’ Basket of Journals as the first paper was published in 2010. The objective of this literature review was to identify and categorize the scope of analytical measures and

A very early and prominent network measure is the strength of interpersonal ties. By examining the retweet behavior of Twitter users, Shi et al. [39] build on Friedkin [40] and analyze the influence of tie strength in the content sharing process. They show that, on Twitter, there is a higher probability of content being passed on through weak ties [39], which supports Granovetter’s [41] weak tie theory. Having analyzed music listening behavior in an online music network, Garg et al. [42] provide empirical evidence that even a network with extremely weak ties can support information discovery. In addition to tie properties, academics assess network structures and their influence on information dissemination. For example, Lerman and Ghosh [15] compare the speed of information dissemination in different types of public social networks by monitoring user activities and hashtags. Using a similar approach, Rios et al. [43] add semantic technologies (topic-based text mining) to SNA, and thus identify topic clusters with common interests. In combination with SNA metrics, like density or HITS, this provides a better understanding of network structures because it allows for identifying, for example, subcommunities or key members. Network structures may change constantly or only exist for a short period [44]. Therefore, another major interest lies in understanding network dynamics. Moody et al. [19] propose different types of visual representation (network flip books and dynamic movies) as a first step to understanding these dynamics. Trier [25] uses longitudinal data of relational events to understand the evolution of networks. Taking a different approach, Quintane et al. [44] focus on network nodes’ reciprocity in order to analyze usage patterns and their impact on geographically dispersed project teams’ effectiveness and duration. More recently, the sentiment of a message has gained considerable research attention. It can affect messages’ dissemination and is therefore analyzed in several ways. By extracting the sentiments of politically related tweets, Stieglitz and Dang-Xuan [45] show how users’ emotions affect the speed and amount of information sharing. Emotionally loaded tweets result in more and faster retweeting. Hillmann and Trier [18] focus on the topographical dissemination patterns of sentiment-loaded postings in several online networks. They show that reciprocity and hierarchy affect the network dynamics and that an early contribution to the network leads to greater influence. Oh et al. [17] analyze tweet sentiment regarding the cause of a crisis situation to evaluate rumor causing factors. By analyzing the content of tweets and the spread of rumors, they indicate that information with no clear source is the most important rumor causing factor. Chau and Xu [14] present a combination of opinion mining and network analysis. By extracting the sentiments of blog posts and analyzing the centralities of bloggers, they show that a

1 Forward search: A search for literature citing the sources already found. Backward search: An analysis of the literature used in the sources already found.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

4

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

high degree centrality is more beneficial for disseminating information and influencing others than a high closeness or betweenness centrality. Even semantic technologies are applied to analyze user-generated content. GarciaCrespo et al. [46] show that natural language processing and the application of semantics can be applied to user comments to categorize user opinions. Networks are an essential part of organizations, and are related to organizational measures, such as employee performance and strategic networking behavior, but also company effectiveness and innovativeness [47,48]. Chui et al. [49] show that the use of an ESN can lead to a 20% increase in employee productivity, increased employee motivation, and improved communication and collaboration. Furthermore, the use of an ESN and the resulting change in information diversity and in communication behavior can transform employees’ network positions [5], thereby affecting their performance [21], or IS proficiency [50]. However, it has been argued that

ESN not only reduce the variety of real life relations due to technical limitations, but also provide its users with a high degree of transparency regarding the network structure and users’ network position, which allows them to use, or even change, these structures according to their interests [10]. By analyzing the revision history of Chinese Wikipedia articles, Zhang and Wang [23] show that editors’ network position influences their contribution behavior in various ways. In this context, Majchrzak et al. [32] argue that others’ visible contribution to the process of knowledge exchange also impacts user behavior. The effect of social technology on employee performance can also be analyzed by distinguishing between online and offline ties. Zhang and Venkatesh [20] use questionnaires and network analysis to show that online communication complements rather than replaces offline communication. To this end, they examine a variety of network characteristics, including centrality measures, tie strength, and network reach, as described in Scott

Table 1 Overview of studies on social networks. Measures

Methods

Data sources

Reference

Influence of tie strength on information dissemination Empirical measurement of information diffusion Effect of network structures on the speed of information dissemination Applicability of topic models to enhance network analysis Different ways to present network dynamics

SNA

Popular tweets

[39]

Empirical analysis SNA; Web analytics; Information retrieval Information extraction; SNA; Web analytics Dynamic network analysis

Logfiles of Music-listening behavior Retweets; User votes

[42] [15] [43]

Importance of time-based events to enhance network analysis Influence of network structures on stability of organizational networks Effect of network structures on information dissemination Information flow and social contagion in crisis communication Effect of network dynamics on sentiment dissemination

SNA; Temporal analysis

Discussion forum; Interviews addressing experiences with the software Random network structures; Friend ranking; Observations of social interactions E-mail communication

SNA, with focus on time and location Opinion analysis; SNA

E-mail communication

[44]

Blog posts; Subscription and comment links

[14]

SNA; Content analysis

Tweets for crisis communication

[17]

Opinion mining; SNA

[18]

Effect of sentiments on user sharing behavior Identification of emotions

Sentiment analysis; SNA Sentiment analysis; Semantic analysis Questionnaire; SNA SNA; Content analysis

Content from  discussion forums  Internet relay chats  Microblogs  Newsgroups Tweets on politics User comments Questionnaire results (focused on relations) E-mails; Calendar events; Instant messages

[20] [21]

Revisions of Wikipedia articles Survey data on  Trust  Employee communication Bookmarks Tags Bookmark access Survey data on interaction  Frequency  Depth E-mails Phone calls SMS Interview data;Content from  Intranet (also wiki)  Project management and tracking tools  E-mail Location-based ‘‘check-in’’ data

[23] [24]

Effect of social software on employee performance Influence of network position on employee performance Effect of network structure on contribution behavior Correlation between trust, communication, and member performance in virtual teams

SNA SNA

Benefits of social bookmarks to increase employee innovativeness

SNA

Effect of network positions on IS proficiency

SNA

Significance of communication patterns for hierarchy detection

SNA

Effect of network structures on idea dissemination

Interviews; SNA

Influence of location-based data on personal decisions

SNA

[19] [25]

[45] [46]

[48]

[50]

[22]

[16]

[59]

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

[51]. Other scholars analyze trust networks and other forms of informal organizational structures. Stepping away from digital traces, Sarker et al. [24] conduct a survey in virtually close, but geographically dispersed, teams. They consequently emphasize the importance of trust in ensuring teams’ success and good performance. ESN also allow researchers as well as managers to identify experts or informal knowledge networks [52] and to access resources, for example, social capital [53]. McPhee and Zaug [54] present a new perspective on social networks and argue that only communication constitutes organizations. Following this view, the type of communication in an ESN also has a large impact on the type of organization [55,32]. Hierarchy has an essential influence on network structures in organizations [56,57]. By means of a network analysis of company communication data, Wang et al. [22] show that data, like phone logs or email traffic, indicates how employees use different communication channels. Building on this, they propose a method to derive a person’s organizational hierarchy by introducing an own metric called HumanRank, which draws on Google’s PageRank [58]. Members of a network also have a mutual social influence, which is another variable that is frequently investigated. To identify influential network members, Ciriello et al. [16] evaluate the diffusion of ideas in online social networks by combining interviews to reveal individual mental models/perceptions of collaboration structures and SNA centrality metrics. To analyze the influence of friends, Shi and Whinston [59] use location-based data of virtual ‘‘check-ins.’’ They show that the likelihood of visiting a new location is not positively associated with the amount of checked-in friends. In other words, many online friends visiting a site does not automatically mean that the related person will also visit it. In summary, our review reveals a great variety of research measures, surpassed by an even greater diversity of data sources used to analyze these objects. Table 1 provides an overview of the (1) variety of research objects, (2) the used analytical methods, and (3) the various used data sources, which describe the major domain of the data rather than the underlying data structure. 2.2. Framework of ESN data dimensions The literature review supports our conviction that research on online social networks is marked by a wide variety of measures, elicited through different methods of inquiry that focus on an even wider variety of data sources. To attain more systematic insights, we developed a framework of data dimensions that can contribute to structuring and harmonizing the increasing research on this domain. The recent work of Chen et al. [60], who define data, text, web, network, and mobile analytics as well as their key characteristics as the major research areas in business intelligence, inspired our approach. An extensive literature review also led to these categories and built a valid foundation for further research. Similar to Chen et al.’s [60] approach and basic categorization, we primarily try to structure the domain of ESN research in terms of data

5

dimensions. We therefore deduced the following four main data categories from the previously mentioned literature review. (1) (2) (3) (4)

Activities (usage data). Content (user-generated data). Relations (structural data). Experiences (reported data).

We subsequently discuss these main data dimensions in more detail. 2.2.1. Activities As soon as someone uses an ESN, a digital trace is left behind. Almost all functions within such a system create usage data, which is collected by cookies, log files, page tagging, web beacons, and packet sniffing [61]. Alternatively, direct access to the underlying databases can facilitate its export. Exemplary data elements include new blog posts or group subscriptions, friendship requests or followers, and page views or the number of users. This data can be evaluated in order to understand the technologies in use, or to improve the design and usability, to accelerate content updates, or even to improve the website’s performance. Kumar et al. [61] and Wang and Lantzy [62] use a combination of certain usage data, like content volume, traffic, responsiveness, and interactivity, to analyze a community’s health. Various web analytics tools can analyze and present usage data [63] and almost all platforms provide a dashboard functionality for easy data access. A significant advantage is that this data can be exported at any time and across any timeframe (depending on the platform settings). However, organizational events and constraints, for example community managers’ promotion of certain topics, or the presentation of important organizational communiqué, can influence this data indirectly. 2.2.2. Content Although usage data provides information on the amount and distribution of activity on a given website, it allows for hardly any conclusions on why users have visited the platform, whether they have achieved their goals, and if the platform was useful in this respect. In order to better understand and interpret these activities, it is helpful to take a closer look at user-generated data, like status updates, blog posts, comments, or tasks and events. Researchers and practitioners can use various methods to investigate these aspects, including content analysis, sentiment analysis, text mining, or genre analysis. Content analysis tries to draw replicable conclusions from textual data to its context [64], or in other words ‘‘Who says what, to whom, why, to what extent and with what effect?’’ [65]. Sentiment analysis aims to determine a person’s attitude toward a specific topic [66]. Text mining allows for discovering new information in written communication, or finding patterns across datasets [67] by using linguistic, statistical, and machine learning techniques. In a genre analysis, one or more communication genres are manually assigned to each content element corresponding to their purpose. Communication genres are ‘‘socially recognized types of communicative actions [. . .] which are habitually

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

6

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

executed by members of a community and which pursue a defined social goal’’ [68]. They are therefore well suited to describe communication practices and usage patterns within groups. Based on Yates et al.’s [68] approach, Riemer and Richter [69] conducted a genre analysis of an enterprise social network. They found that microblogging in the enterprise context differed greatly from that in the private context. It was more work-related than some upper managers expected. Unwanted behavior, like non-workrelated chatter, does not automatically emerge when social technologies are applied in the enterprise environment. It is quite simple to export the primary material for analyzing content. Some platforms even offer features like RSS-feeds to facilitate this. However, the question of privacy arises, which needs to be addressed beforehand for example in terms of work council agreements, or anonymization methods. Furthermore, depending on the concrete analysis method, subjective bias is another issue to consider. 2.2.3. Relations When users interact with each other in an ESN, they create connections. The study of relations between persons, or social network analysis (SNA), has a long history in the social sciences [70]. However, in the social sciences, these relations were often created manually, for example, by means of interviews, which is a painstakingly slow process. If an ESN is used, connections of various kinds occur automatically when users interact on the platform and these are documented automatically [11]. Connections can include following other users and/or commenting on, liking, rating, and/or sharing their contributions (depending on which mechanisms are available on the platform). Consequently, connections emerge between people, between people and content, or even directly between content elements. Various metrics can be used to analyze the structural data [51], as outlined in Section 2. It can further be presented as a graph, which is a set of nodes that

represents the network actors, and a set of edges (or links or ties) representing the relationships between these actors [51]. ESN often provide several types of content and interaction features. Even though they can be exported quite simply by directly accessing the database, the major challenge lies in the analysis itself. Researchers need to decide if and how to combine the exported data in one network, or if it should be treated separately. Furthermore, as Kane et al. [10] show, researches need to consider the platform-intended tie design and platform features when interpreting the results. 2.2.4. Experiences Users’ experience of and the attitude they develop when using a platform provide insights into their experiential life [71]. Interviews or questionnaires can be used to collect these qualitative user perceptions. Our review shows that interviews or questionnaires were used to analyze the dissemination of ideas [16], the motivation to share knowledge [6], or social software’s effect on employee performance [20]. Given all the steps required from the selection of interview partners and the coordination of appointments to actually conducting interviews and undertaking the final analysis [72], the collection of experiences requires a huge effort and lots of resources. However it still only shows a snippet of all users’ experiences [11]. Furthermore, the design of interview guidelines or questionnaires can influence the results. Hence, researchers must consider all these aspects when planning an analysis. In addressing our first research question (RQ1), we integrate the metrics used in our reviewed studies and the measurement domains of the business intelligence approaches [60], as shown in Table 2. The above framework of data dimensions provides an overview of the main data sources for a mixed methods inquiry. Extending beyond existing academic contributions,

Table 2 Proposed data category framework. Data dimensions Activities

Content

Relations

Experiences

Characteristics

 Data can originate from various sources  The automatic collection results in an extensive amount of data  The quantity and quality of an analysis depend on the features of the underlying system, the export options, and the complexity of the database structure  Is documented in a corresponding context / related communication (threads)  Requires manual preparation (e.g., data selection or pre-processing)  Can be analyzed manually or automatically  Allows for subjective bias in manual analysis  Supports multilingual settings  Different types of relations  Are affected by platform features and intended tie design  Allow for the analysis of network dynamics and evolution  Must be documented manually and sporadically  Can be assessed on a sample basis only  Allow for a deeper understanding of user intensions and the underlying context  Require a huge effort to prepare, conduct, and analyze  Allow for subjective bias in the collection and analysis process

Exemplary data collection methods

Exemplary data analysis methods

Exemplary studies

Log files; Tracking pixel; cookies

Web analytics

[15,43]

Structured content export

Genre analysis; Sentiment analysis; Text mining; Content analysis

[14,18]

Structured content export

SNA; Dynamic network analysis

[25,19,20]

Interviews; Questionnaires

Content analysis

[16,6,44]

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

we now address our second research question and assess the implications of using these data dimensions and combinations in an empirical ESN case study. This objective suggests a mixed methods research approach, which is characterized by a combination of qualitative and quantitative analyses. This method yields additional insights and improves the validity of the results by means of triangulation, that is, a systematic comparison of results from different sources of evidence [73,29]. Triangulation not only improves researchers’ confidence in their findings, it also yields valuable insights into the inadequacies, gaps, biases, and white spots of inquiries that only focus on one data dimension [35]. For example, by combining content analysis and network analysis, we have a better chance of understanding why an online network developed a particular structure [74]. As motivated in the introduction, we aim to go beyond the development of a theoretical research framework. We therefore proceed to assess and apply our framework to a comprehensive empirical case study of organizational online social networking. This serves as a proof of concept, but also enables us to focus on a careful methodological reflection. More specifically, we want to identify how different methods overlap or inform each other. We also expect conflicts to arise when comparing interpretations from different data sources. Subsequently, our case results help to inform the development of a systematic mixed methods research approach to ESN studies.

3. Case study Our selected case organization is the medical service unit of the German Armed Forces (Deutsche Bundeswehr). It employs, amongst others, 2700 medical officers and 1600 trainee medical officers assigned to military medicine, military pharmacy, veterinary medicine, or dental medicine. Contrary to most officers who study at one of the two Bundeswehr universities in Munich and Hamburg, these employees study medicine at civil universities across Germany. The workforce and the students are distributed across five hospitals, 37 universities, and 200 other facilities. The organization was selected for its strong hierarchical culture, which allowed us to focus on online networking across hierarchies. In this controversial and contextdependent domain, we expect each data dimension to yield different results due to their varying ability to capture user intentions, planned versus actual user behavior, and the organization’s contextual influences. For example, we consider interviews to more accurately capture perceived contextual norms, as well as how they influence behavioral intentions (why something is done or not). On the other hand, structural SNA usually focuses on manifest content, which means it is unable to indicate the ‘‘invisible’’ context that might explain why certain online networking behavior is observable (or not). By carving out these differences in the methods, we aim to gain better insights into the inconsistencies or mutual validation of different data dimensions. This focus on the implications of data collection methods also implies we

7

do not primarily intend to explain hierarchy formation in this paper. The relevant period for our research starts in 2009, when the department responsible for the training of medical officers identified a need for a platform to support knowledge transfer and collaborative learning. Trainee medical officers specifically expected state-of-the art communication tools. The department decided to implement an ESN, which we refer to as Med-Net in this paper. The main goals of the Med-Net project are to (1) foster knowledge transfer and collaborative learning among staff, (2) improve the quality of education and the in-service training of new employees, (3) strengthen the corporate identity and networking of staff, and (4) create a collaborative knowledge base [75]. Med-Net was initially intended for trainee medical officers, but was extended to all medical staff later on. Even though the use of Med-Net is voluntary, superior officers strongly encouraged its use. These characteristics were considered in the analysis. The project started in November 2010 as a pilot, which was developed and maintained by our research group. It was handed over to the Bundeswehr’s medical service in 2013, which has since then been running the platform. In the course of this handover, we had the opportunity to evaluate Med-Net, which gave us exclusive access to an otherwise inaccessible data set. The goal of the evaluation was to document existing use cases and to evaluate their usage, as well as to identify best practices. This broad objective made this case very suitable for our research objective, as we could carry out a very comprehensive analysis of all the domains proposed in the above framework (Table 2). 4. Methods and data collection In the following sections, we describe the methods used in the case study. The theory section motivates these methods, which are based on the framework in Table 2. Later, we describe in detail the data dimensions used and how they were collected in our case setting. 4.1. Method combination Several studies have argued that the ways in which ESN are used are not prescribed in advance and that the users themselves decide how to apply such software and the given features by incorporating ESN into their daily tasks [76,10]. Since use cases are not defined, evaluations are a comprehensive and exploratory inquiry and allow for a mixed methods approach to be used to evaluate a social network case. The research thus involved multiple methods, which helped verify one method’s results by means of those of another method [34]. A mixed methods approach can also help expand the results to enhance the integrity of the findings [33]. Creswell and Plano Clark [33] present six major mixed methods designs. They differ in order and in their dependency of the qualitative and quantitative analysis phases. We used a convergent parallel design, which is suitable to develop a complete understanding of the domain of

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

8

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

interest [33]. Other designs are more suitable when the focus is on certain aspects. Hence, our qualitative and quantitative analyses are based on data from the same period. Neither strand influenced the other and were handled the same. After the collection, the results were compared and interpreted. Following the methodological overview suggested in [60], our quantitative data collection and analysis consist of web analytics and a quantitative social network analysis. The qualitative strand consists of expert discussions, genre analysis, and semi structured interviews. According to Venkatesh et al. [34], qualitative methods are suitable for six aspects of social network analysis: the exploration of networks, the examination of network practices, network interpretations, network effects, and network dynamics, as well as access to actors and networks. Building on this framework, we mainly focused on the exploration of the network, the analysis of network practices, and on network interpretation. The exploration of the network forms the basis for further investigations. We gathered preliminary information about the network by means of discussions with experts. In this case, the experts were the persons responsible for the community management and technical development of Med-Net. This gave us a deeper understanding of use cases and of the platform’s functionalities. It also helped us understand the kinds of users present in the network. The exploration moreover allowed us to focus further research and interpret the quantitative findings later on. The examination of the social practices was done by means of semi-structured interviews and a genre analysis. This helped us reconstruct and understand the specific user practices. To better interpret our findings on the relevance of the network for its users, we combined the qualitative data with the results from our quantitative analysis. Our approach allowed us to analyze the dissemination, the conditions, and the consequences of usage patterns.

genre analysis and manually processed, as Yates et al. [68] suggest. Furthermore, after a process of anonymization, we analyzed the relations of all the registered Med-Net members. We assessed three different types of relations, which were based on the following Med-Net technical features: (1) content and comments, (2) contact requests, and (3) direct messages. On the content and comments level, nodes are defined as users who create a content element (either an initial content or a comment). Edges were defined as comments on initial content. In the case of contact requests, edges comprise accepted contact requests. Nodes are senders and recipients of these requests. When direct messages are examined, nodes are senders and recipients of direct messages, and edges are created by sending a direct message to a user via Med-Net. During the analysis, we considered these three dimensions separately and did not merge them into one network. A social network analysis was performed on the data and the result was presented in a graph created using Commetrix2 software. We conducted semi-structured interviews to gain an understanding of the users’ experiences. To this end, a guideline containing 13 questions was created. It focused on the activities and position of the interviewees, their experiences with civil social networks, their current MedNet use, and suggestions for optimization. All the interviewees were Med-Net users. They were selected on the basis of their activity index, a Med-Net feature. This index incorporates the number of log-ins and views, as well as the created content and comments. Based on this index, we identified very active and passive users. Furthermore, we considered officers of various ranks and with different tasks at different locations in Germany to gain a broader perspective. In total, 25 users were invited to participate in a 30-min interview. Ultimately, 13 phone or personal interviews were conducted. With the interviewees’ consent, the interviews were recorded, anonymized, and transcribed. After the evaluation, the records were deleted.

4.2. Data collection, preparation, and analysis

5. Analysis results

Activities can be collected via cookies, log files, page tagging, web beacons, and packet sniffing [61]. In our case, the underlying content management system, Drupal, automatically captured the data, which we exported to Excel files and analyzed manually. Owing to Drupal’s specifics, we could easily export statistics on, for example, visits, views, and the number of content elements. However, other desired information could not be accessed, for example, details of search requests and poll data. We did not consider data from the first year, as this was a testing phase and we believed this data would distort the outcome. To examine communication in groups, we conducted a genre analysis of the content of selected groups in Drupal. The Med-Net groups were manually ranked according to their number of articles, members, and views. Seven of the top ten groups were selected based on their activity and privacy setting. Only public groups were analyzed. All 1155 content elements (status messages and articles) from the seven selected groups were exported for the

This section presents selected results of our mixed methods approach, using our proposed data dimensions as the main structure. Elements are selected for presentation if they serve to illustrate the interplay between the different methods and their results. 5.1. Activities At the time of the data extraction, more than 1400 users had registered on the platform. Although many only passively followed the conversations on Med-Net, almost a quarter of the users were actively involved. Even when the number of new users decreased, there was a growing interest in the existing content, which the number of visits reflected. The number of visits also showed peaks from time to time. Additionally, the different communication features (content, comments, and private messages) were used very unevenly. Fig. 2 shows that more than half of all the created artifacts were private messages. 2

http://www.commetrix.net/.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

Comments 28% Content 17% Private messages 55%

Events Blogs 28% 33% Survey 3% Arcle 36%

Fig. 2. Usage of Med-Net communication features.

The examination of the content types clearly showed that events played an important role, even though the event feature was not fully implemented at the time. More than 30% of the created content was an ‘‘event’’ type. The analysis also showed that the users primarily created groups that required group membership or invitation. Furthermore, the usage data indicates that a large amount of communication took place within those closed groups. In addition, a ranking was created of the top 10 Med-Net groups. The groups were ranked according to their number of articles, members, and views. Five groups appeared in all three rankings. This gave an indication of the importance of these groups and the topics discussed therein, but also showed their potential allies, or good sources of feedback for the further development of Med-Net, as the members of these groups are very active, can provide good feedback, and may work as information hubs in the future.

5.2. Content The genre analysis revealed that Med-Net was mainly used for coordinating events, appointments, and meetings. More than 40% of all articles and status messages are event related, as can be seen in Fig. 3 (a detailed description of the genres can be found in Appendix A). This corroborates the usage data analysis’s results. The genre analysis furthermore shows that, besides coordination, the discussion

9

of opinions and sharing of interesting information was a dominant aspect of the communication. The analysis also revealed a use case that was not previously obvious: the selling of professional literature. While this genre is not particularly remarkable, it highlights that users follow their individual needs and, in effect, create new uses. This would not have been recognized if only usage data had been analyzed. The interview data did not suggest this use case, as we were unable to interview all the Med-Net users. We thus emphasize that a combination of methods can compensate for the disadvantages of individual forms of inquiry. Praising other users, or thanking them, was quite common. This was very positive behavior and contributed greatly to motivating the users. They recognized that their commitment was valuable and that they could help others. Additionally, the users had established and maintained a polite interaction culture, which was inferred from the salutations and forms of greeting. The interviews confirmed these findings. None of the other applied methods revealed such behavior. The genre analysis moreover showed that classified documents were not often exchanged, even though this was allowed to a certain extent. 5.3. Relations Med-Net’s entire network structure was analyzed on the basis of the following different network types: (1) content and comment, (2) direct messages, and (3) contact requests. Furthermore, we analyzed the seven groups, which we initially selected for the genre analysis, in more detail. Our goal was to understand how the hierarchy of the federal armed forces would influence their communication and networking behavior. Hence, the users of these groups were grouped according to their military rank. Table 3 provides an overview of the total number of active soldiers with their respective ranks within the selected groups and overall. We used the NATO ranking code [77] for military ranks. The grouping of the ranks in Table 3 (e.g., OF3-5) follows the common Bundeswehr structure. Because not every member of the medical service is a Med-Net member, the distribution of users across ranks

Fig. 3. Overview of identified communication genres.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

10

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

Table 3 Overview of Med-Net users according to a relational analysis. Military rank

Number of soldiers (in Med-Net)

OF3-5 OF2 OF1 OR 5-9 OR 1-4 Civilian/n-a

158 76 665 329 89 82

Number of soldiers in selected groups 23 11 97 48 13 12

in Med-Net does not correspond to the distribution in medical service in general. 5.3.1. Content and comments Even though 388 users contributed to the content, the majority of it was derived from only a few people, who constitute the core of the network. However, these core users do not form a dense network themselves. Furthermore, the analysis shows that the majority of the core users have a high military rank. Med-Net is not used for organizing military matters. Hence, no tasks or commands are given to subordinates via this platform which was confirmed by the genre analysis. Therefore, this distribution indicates the higher ranks’ willingness to share their knowledge, and offers the lower ranks a good opportunity to profit from it. The average degree of connectedness is very low. In addition, we detected 54 isolates, that is, users who published content but received no reaction.

connected than their content and comments or direct messages indicate. A comparison of the three distinct network types shows that lots of users know each other and stay in close contact. However, the majority prefers to communicate directly and privately instead of openly (Fig. 4).

5.3.4. Hierarchy Besides these three levels of analysis, we were interested in the military hierarchy’s influence on the communication and network behavior. We also wanted to double-check interview statements referring to the ranks not playing an important role in Med-Net. The analysis, based on the content and comments of the seven selected groups, revealed that this assumption was not true, which Fig. 5 shows. Connections across ranks mostly occurred across the higher ranks. However, the majority of the networking occurred within the same hierarchy level, especially between OF-1s. Comprising 48%, OF-1s were also the biggest group of users and provided most of the content. Although unsurprising, it is striking that OF-1s bridged the structural gap [78] between the lower and upper ranks. There was therefore a strong relationship between OF-1s and OR-5-9s, as well as between OF-1s and OF-3s.

5.3.2. Direct messages Similar to communication via content and comments, the network of direct messages also shows a small number of heavy users. They can be regarded as superspreaders, that is, people with many contacts and a large content output. 5.3.3. Contact requests The graph of this network differs very strongly from the other two. The average degree of connectedness is much higher and the active core is much larger. Furthermore, this network covers the majority of the Med-Net users. In sum, there are 1192 nodes and 4375 edges. In total, 71% of the users have at least one approved contact. Based on contact requests, Med-Net users are far more

Fig. 5. Network structure based on military ranks and on locations.

Fig. 4. Graphs of three network types (1: Content; 2: Direct Messages; 3: Contact request).

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

5.4. Experiences Besides an analysis of the numbers and content, we also conducted an interview study. This study was centered on the following questions: 1. How is Med-Net currently being used? 2. Have users established new ways to use it? 3. How can Med-Net be improved? The motivation to use Med-Net and how it is used depend largely on the user’s rank and role. The interviews, for instance, show that, from time to time, there were events at different locations during which Med-Net was presented. After such events, superior officers occasionally urged the lower ranks to use Med-Net. The officers in charge of caring for lower ranked officers mostly used it as a fast and barrier-free way to make contact with officers with whom they were not close. The trainees regarded obtaining the latest information on military and academic issues as Med-Net’s major benefit. Across all ranks and positions, the interviewees regarded Med-Net as a good source of forms, regulations, and other documents. But they were also uncertain whether Med-Net should be used to exchange these documents. The interviewees emphasized several distinct Med-Net features, for example, the virtual hospital and the location profiles. When asked if they used Med-Net in other ways than initially intended, none of the interviewees mentioned a new use case. All the interviewees stated that military hierarchy does not play an important role in the use of Med-Net. The threshold for making direct contact with a superior officer through Med-Net was described as very low compared to the official channels. Some interviewees, however, feared that this reduced distance could lead to a loss of respect. But when asked if this actually happened, no one confirmed it, and many mentioned that communication was polite and respectful. Some suggested to improve the usability in certain Med-Net areas. Furthermore, the search functionality was said to have potential for improvement. The users also mentioned that Med-Net did not offer a possibility for quick appraisals – for instance a ‘‘like’’ button. 6. Discussion Based on our empirical inquiry, we will now refer to our initial research questions and discuss the respective results. 6.1. Data dimensions in ESN research To address the first research question, we developed a framework of available data dimensions in the ESN context. Our case study showed that these four dimensions provide a comprehensive view of ESN use. However, this comprehensiveness could result in getting lost in the data. Therefore, these dimensions should be seen as the top layer only. Each data dimension should be divided into meaningful sub-categories to focus future research. For example, activities can contain statistics about the content elements, or the users. Content can be used, for example, for information extraction, topic tracking, summarization, categorization, and more [79]. Future research should

11

determine which of these data mining technologies are valuable in the context of an ESN analysis. Furthermore, due to its technical features, ESN allow for analyzing multiple types of relations. Researchers need to consider whether and how to integrate relations, such as followers, friendships, direct messages, or even user comments. Assessing the experiences of ESN users can generate a deeply contextual understanding of users’ outer and inner worlds [71], producing a wide variety of results. We therefore maintain that a standardized set of interview or survey questions will contribute comparable results in the analysis of a specific ESN, or across different systems, or different case studies. Consequently, further research is required to describe each dimension separately in order to gain an applicable framework. 6.2. Mixed methods approach in ESN research Based on our case analysis, we next address our second research question and discuss the applicability and possible benefits of a mixed methods approach in analyzing an ESN. Our exploratory case study suggests that individual analytical methods for examining online social networks can improve our academic understanding, but they also provide a basis for managerial interventions. An analysis of activities can reveal areas of high or low contribution and interest in an ESN. Companies can use this data to identify popular features, or content types. Content analysis results are, for example, usage patterns or hot topics. The interview data shows the underlying motivation for the observed behavior, as well as how individuals perceive the ESN’s benefits and limitations. A network analysis reveals how the network is constituted and shows the structural configurations that influence communication. All this data, even if considered separately, can already help inform the development of communication strategies, the improvement of workflows, and the platform development in terms of its usability, technical features, and social structure. Furthermore, such data provides a basis for deducing arguments to justify the ESN’s relevance for the management, or to convince new users to register. However, combining different analytical perspectives enables a better account and explanation of the underlying phenomena. Mixed methods analysis moreover yields more robust results as different methods lead to consistent findings that confirm each other. In the next section, we exemplarily show how this relates to our case study, that is, we discuss the benefits of a mixed methods approach, but also its limitations. Based on this, we develop guidelines for the separate application of the data dimensions, as well as in the context of mixed methods ESN analysis. 6.3. Benefits and limitations of mixed methods research of ESN In some cases, our approach revealed contradictions. For example, we found that users’ perceptions of their network can strongly differ from the measured log data. These contrasts cast doubt on studies that focus solely on surveys or interviews. Our data clearly shows that valuable insights

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

12

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

might remain undetected. In more detail, our adopted mix of methods specifically highlights the following four situations. 6.3.1. Mutual confirmation of results The quantitative activity analysis helped identify coordination as a main reason for using the network. The results show that the qualitative results emerging from the genre analysis and interviews support this log-filebased interpretation. The interviews also emphasize the polite communication style, which was first indicated by the genre analysis, even though the interviewees were not directly asked about this. Furthermore, our quantitative results from the network analysis show a high degree of connectedness. This is consistent with the interviewees’ perception that the social network is beneficial in terms of making contact and staying in contact with remote actors. Consequently, the main use objectives and individual advantages of online social networking emerged consistently in different types of inquiry. 6.3.2. Conflicting results The mix of methods also yielded conflicting results, especially in the context of the social desirability of the outcome. For example, the interviewees claimed that hierarchy does not play an important role in communication on the network, which would be in line with an ideal group culture. Conversely, the SNA showed that networking is mostly done within the same hierarchical level, which indicates that the organizational hierarchy affects the communication behavior. People might therefore be inclined to prefer the same hierarchical level, but would in principle be ready to collaborate more across levels. The interviews thus emphasized the potential scope of acceptable use rather than highlighting the typical cases. The utilization of document exchange is a similar example. The view the users expressed in the interviews differs from the usergenerated data. The interviewees mentioned that the network is a good place to exchange (classified) documents. However, this behavior was not detected in the genre analysis. We did not track the number of uploads or downloads in the usage data analysis, which could have made this result even more significant. 6.3.3. Clarification of observed phenomena Some results could initially not be explained on the basis of just one type of inquiry. For example, our SNA pointed to an uneven distribution of users across different locations. Only the interviews revealed the context, that is, that network introduction events only occurred in major locations, which enabled a useful interpretation of the SNA results. As a background story, the interview-based narration of the introduction events also helped explain the peaks that the activity analysis identified in the number of visits. In these events, the superior officers occasionally urged the lower ranks to use Med-Net. As a result, the number of visits increased right after the event. Owing to the special handling of the registration process, the number of registrations did not increase. The usage data also

showed low use of certain platform features. In this context, the interviews revealed that one reason for this was the partially poor usability, with the interviewees describing a confusing platform layout in some Med-Net areas. In all of these cases, the qualitative data helped highlight the context, thus making the quantitative results more comprehensible. 6.3.4. Revelation of new insights A mixed methods approach is more likely to identify new insights. A new use case (‘‘offer literature for sale’’) was discovered via the genre analysis, while no other method indicated this. Interview results might be insufficient here if the selected interviewees do not use the ESN in that way. Another example is the combination of SNA and web analytics, which revealed unexpected user behavior. It showed that the heavy use of direct messages was limited to a small group of users. Even though the genre analysis indicated that users talk about the ESN, no direct benefits were mentioned. Only the interviews revealed new qualitative benefits, besides the already mentioned networking benefit. Finally, the combination of web analytics and genre analysis revealed the necessity to review and promote important content: in our case, the usage guidelines. Both perspectives suggest that the users neither read the usage guidelines (web analytics), nor talked about, or referred to them (genre analysis). To sum up, this shows that the strengths of some methods can compensate for the weaknesses of others. Table 4 shows a structured summary of all the aspects described above. In most of the issues shown above, interviews played an important role and strongly contributed to a better understanding by describing the context. This revealed several issues, which future studies can explore and further develop. In our case, the mixed methods approach had a convergent parallel design, which allowed us to compare and triangulate certain data. But, as indicated in the results section, even more results or more significant results would have been possible with other configurations of mixed methods. We therefore suggest taking an exploratory sequential design, in which the qualitative data collection and analysis build on a previously conducted quantitative data analysis. We moreover suggest using a multiphase design. Here, a first qualitative (pre)study informs a subsequent quantitative (pre)study. This in turn informs the methods that are used in the mixed methods study [33]. This will allow for a step-by-step evaluation of different aspects, building results and insights on top of one another and making use of the four aspects described in Table 4. 6.4. Guidelines While using the four data dimensions in a mixed methods approach, we faced some methodical and practical challenges. We now address these challenges by providing guidelines for an individual application of the data dimensions, as well as in the context of the four identified aspects of mixed methods ESN analysis (see Table 5). Because the guidelines are based on our experiences in the course of

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

13

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx Table 4 Overview of possible mixed methods advantages. Example

Data dimensions

Details

Mutual confirmation of results Confirmation of intense coordination use

Activities Content Experiences

Large number of event content types Large number of coordination-related communication Summary: Med-Net is good for coordination

Confirmation of communication style

Content Experiences

High use of polite salutations and greetings Summary: Respectful communication takes place

Confirmation of benefits of Med-Net usage

Relations Experiences

High degree of connectedness based on contact requests Summary: Med-Net is a good way of staying connected with remote officers

Conflicting results Effect of hierarchy

Interviews Relations

Summary: Hierarchy does not affect communication Mostly connections within the same hierarchy

Usage for document exchange

Experiences Content

Summary: Med-Net is a good place for document exchange Missing genre ‘‘document exchange’’

Clarification of observed phenomena Uneven distribution of users across different locations

Relations Experiences

Detection of uneven distribution Clarification: Introduction events mostly in Munich and Hamburg

Visit peaks

Activities Experiences

Detection of peaks Clarification: Higher officers command participation at introduction events

Low use of certain features

Activities Experiences Content

Detection of low usage Clarification: Lack of usability (user-unfriendly) Missing genre ‘‘document exchange’’

New use cases (book sale)

Activities Experiences Relations Content

No indication No mentioning of this use case Not suitable for such results Revelation of entries about selling books

Imbalanced communication via direct messages

Activities Relations

Heavy use of direct messages Heavy usage only within a small group of users

Qualitative benefits of Med-Net

Activities Content Relations Experiences

Not suitable to show qualitative results No indication visible Not suitable to show qualitative results Summary: Med-Net helps with daily routines and improves knowledge exchange.

Better promotion and review of usage guidelines necessary

Activities

Low interest in usage guideline pages (based on number of views)

Content

No mention of usage guidelines in the communication

Revelation of new insights

Table 5 Guidelines for a mixed methods analysis of ESN. Data dim.

Research guidelines

Activities

Individual application  Due to database complexity, consider extra efforts to export data to the required structure  Consider the volatility of events that are typically subject to various degrees of importance over time (peak periods)  Limit complexity by defining subjects of interest and time periods in advance  Detecting main event types can help researchers focus on the dominant use cases Confirmation  Corroborate usage behavior cases by triangulating the results from reported data (experiences) Conflicts  None identified Extension/revelation  Use activities as indicators of issues that require a deeper analysis of other data dimensions  Consider network structure when analyzing the usage intensity of platform features to obtain a clearer picture  Consider the context (from experiences) for a better understanding of possible activity peaks or other significant usage changes (continued on next page)

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

14

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

Table 5 (continued) Data dim. Content

Relations

Experiences

Research guidelines Individual application  Consider the context of conversation for adequate coding results (communication threads)  Consider possible multi-lingual environments and their impact (e.g., a discussion of complex issues may be more likely in users’ mother language) Confirmation  Select a broad variety of content to allow for confirmative effects Conflicts  Look for missing genres or usage patterns by comparing the results of other data dimensions Extension/revelation  Look for new genres or usage patterns by comparing the results of other data dimensions Individual application  Define focal relationship(s), for example, consider combining different relational data  Consider the restrictions and affordances of the platform in the interpretation of the data (some relationships are easier to create than others)  Consider the design of possible relations as predefined by the platform Confirmation  Network visualizations and centrality metrics can confirm personal perceptions of being connected as expressed in interviews Conflicts  Consider that network analysis may provide a biased/different view, as one relationship is usually in focus, whereas interviewees might be able to switch their referenced relation quickly and subconsciously  Consider that, in contrast to experiences, a networking structure excludes the context and the motives behind networking (e.g., plans for cross-hierarchical use) Extension/revelation  Consider interviewees’ different degrees of embeddedness in order to explain different user groups’ different views  Examine whether identified dominant behavior occurs in a strongly connected (smaller) cluster, or is dispersed across the whole network; the interdependency of interviewees’ experiences, and their activity types Individual application  Consider an appropriate interview or survey design Confirmation  Interviews should address aspects of interest (inspired by other data dimensions) explicitly to allow confirmation (or reveal conflicts) Conflicts  Results may represent the planned/intended behavior, while all other dimensions are more likely to assess the actual behavior Extension/revelation  Explicitly ask for own use cases and usage patterns and their reasons in order to understand user intentions  Plan a preparative interview with platform owners to understand the relevant conditions (e.g., the rollout process, registration process, and promotion activities)

the case study, they are neither complete nor described in other studies. 7. Conclusion Communication within a networked organization creates a large variety of data, for example, usage data, usergenerated data, structural data, and user perceptions. The analysis of this data yields valuable insights for relevant business decisions. The applicable analytical methods and the questions that these analyses can answer are equally diverse. In this paper, we proposed a framework of four basic data dimensions and identified their key characteristics. We argued that a mixed methods approach can produce better results than a single method approach, because it considers combinations of different sources of evidence. We explored the application of our data dimension framework and of mixed methods research in a comprehensive and multi-dimensional empirical case study. Based on our analysis, we reflected on four aspects of mixed methods approaches. These are the confirmation of prior results, the clarification of observed phenomena, the contrasting of results, and the discovery of new insights. In conclusion, we explained that a mixed methods

approach is helpful in that it not only shows activities in an ESN, but also the motivations behind them, as well as the resulting social structures. Our case also showed that a mixed methods approach is not only applicable to the evaluation of an ESN for research, but it also creates additional value for organizational decision makers, as the evaluation results are more significant and more comprehensive. Furthermore, the results of the case study, in the form of guidelines, provide a toolbox for practitioners to generate multi-dimensional insights into their company networks, which, in turn, can support better decision making. The selected case revealed that enterprise social networking is a multi-dimensional phenomenon that is reflected in different data dimensions. Researchers need to consider the implications of only using one data dimension for their research. They can also benefit from various combinations of different data dimensions. We further conclude that a strong focus has to be defined when using mixed methods to avoid results that are too broad and to easily establish a link between the different methods. Future research should therefore focus on conceptualizing enterprise or business-related questions and examine how they can be addressed and assessed by means of mixed methods.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

15

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

Appendix A. Genre description

Genre

Description

Announce an event

A future event or meeting is announced This can be a march or target practice, but can be an exam or a leisure time activity as well  On 23 November 2011, there will be target practice from 10 am to 3 pm in [anonymized]. Whoever is interested, please let me know so that I can forward the data. Call me at 0172-[anonymized]  Dear comrades, I have scheduled an appointment to March in the sports area on Friday at 8 am

Accept or reject event invitation

Users respond that they will participate in an event or that they cannot participate

 Count me in as well  Sounds great! I am interested  I am sorry, but I won’t make it. I have to write the second draft of my doctoral thesis until then Share event details

More specific and concrete information regarding an event is being requested, or being stated respectively (e.g., time or place) The event may also be cancelled  Hey guys! Are there more details about the weekend regarding the program (Does the existing program remain unchanged) and dress code (DA, camouflage or even BGA)?  Hi comrades, [. . .] a quick explanation of what this is. The placement test is the test that will get you 30 credit points (you can check out the exact distribution of scores in the following excel sheet) [. . .]

Task coordination

Necessary tasks are stated. However, tasks are not delegated to a group or person directly

 Please read the document completely and carefully and fill it out. Please attach the proof that the university has received the payment to the application Discuss opinions

Share one’s opinion regarding a specific topic and possibly discuss the topic in more detail

 I find it an interesting dispute about leadership and speech, not only for members of the land forces. ‘‘Comments and discussion wanted and needed,’’ in order to stick to the motto of the linked polemic paper Praise other users

Praise the post of another user or mention their engagement in a positive manner

 Thank you for posting this summary! The chat was very interesting and especially the case study at the beginning was quite fascinating  Thank you for helping me make contact! Share status update

A short message is being posted on what the user is doing at the moment or how far he or she has gotten with a specific task or project  I am currently working on planning the course. But it’s not that simple. [. . .]  Status: Begin brainstorming and research. Responsible: [anonymized]

Share own experiences

Users report their own experiences or document results of events in which they participated

 Well, we usually had tutorials 2–5 days before the exam in a ca. 1.5–2 h PowerPoint coaching session. . .  From 27 to 29 November 2012, there was a workshop in the SanAkBw in Munich. Present were amongst others: [. . .]. In work groups, we defined skill profiles for each of the training periods and analyzed the deficits  Summary of the chat with medical officer [anonymized] regarding the topic ‘‘anesthesia and emergency medicine’’ from 23 January 2013. At the beginning, participants were given a case study, which they eagerly solved. In the following [. . .] (continued on next page)

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

16

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

Appendix A (continued) Genre

Description

Share tips and tricks    

Users give tips, which they think might be helpful to other users Emergencies happen without notice, suddenly and surprisingly [. . .] to notify relatives and friends please take down the name and further contact details in the form [. . .] Movie-tip: ‘‘So close to death.’’ Monday, 21 November on ARD at 10:45 pm the documentation: ‘‘So close to death’’ regarding the situation in Afghanistan The covers and content lists of the last letters to members are published in this blog. Every user can search the blog for interesting articles and request a copy of it from our branch office Here you may ask questions that have not been answered in the text or the program; see link

Share a link or reference

   

Users publish a link to a website on Med-Net or externally, which they think might be interesting to other users. The link can also lead to attached files or contain an indirect reference (e.g., naming the place of the content without posting a hyperlink)

The program of the MKP: http://www.[anonymized] Attached, you can find the protocol of the SVV I/2013 [attachment] Further information will follow in the group: FachSanZ Köln-Wahn More detailed information can be found in the group SanOA Rheinland

Ask a question Users have a question and ask other users for help  Is the topic mandatory term still recent? What decisions have been made?  If you come as a backup officer, do you have to participate in the preliminary practice? Clarify an issue

Users explain an already addressed issue in detail. This happens as a response to other user’s requests

Share contact details

Users post their own contact details or these of persons relevant to the addressed issue

 Captain [anonymized] phone: 089-[anonymized] Present a project

Users describe a project in which they are involved or on which they have detailed knowledge  I would like to highlight the project ‘‘SimMed,’’ which was presented to me at a conference. The project just recently won the European eLearning Award. Initial point: student doctors and doctors in further training often have to advise someone, make decisions, and conclude with a proper diagnosis under heavy time pressure in order to save the lives of patients

Offer literature for sale

Users offer own materials, such as specialized literature for sale  Hello, I sell the following books: Explain 7. Edition. Inner medicine [. . .]; 8. Edition surgery [. . .]

Meta-communication regarding Med-Net

Users ask questions on how to use Med-Net or give tips to improve its usability  Even if the conditions of the group structure are not optimal right now, it should fall asleep! Do we have the possibility to upload the compendium in here?

References [1] J. Zhang, Y. Qu, J. Cody, Y. Wu, A case study of micro-blogging in the enterprise: use, value, and related issues, in: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, ACM, New York, NY, USA, 2010, pp. 123–132. [2] A. Majumdar, S. Krishna, P. Bjorn, Managers’ perceptions of social software use in the workplace: identifying the benefits of social software and emerging patterns of its use, in: Proceedings of the 19th Americas Conference on Information Systems (AMCIS), Chicago, Illinois, USA, 2013. [3] S. Aral, C. Dellarocas, D. Godes, Introduction to the special issue— social media and business transformation: a framework for research, Inform. Syst. Res. 24 (2013) 3–13. [4] A. Bharadwaj, El Sawy, A. Omar, P.A. Pavlou, N. Venkatraman, Visions and voices on emerging challenges in digital business strategy, MIS Quart. 37 (2013) 633–634. [5] G. von Krogh, How does social software change knowledge management? Toward a strategic research agenda, J. Strategic Inform. Syst. 21 (2012) 154–164.

[6] M. McLure Wasko, S. Faraj, Why should I share? Examining social capital and knowledge contribution in electronic networks of practice, MIS Quart. 29 (2005) 35–57. [7] D.M. Boyd, N.B. Ellison, Social network sites: definition, history, and scholarship, J. Comput.-Mediated Commun. 13 (2007) 210–230. [8] R. Agarwal, A.K. Gupta, R. Kraut, Editorial overview—the interplay between digital and social networks, Inform. Syst. Res. 19 (2008) 243–252. [9] A. Richter, M. Koch, Functions of social networking services, in: Proceedings of the 8th International Conference on the Design of Cooperative Systems, Carry-le-Rouet, France, 2008, pp. 87–98. [10] G.C. Kane, M. Alavi, G. Labianca, S.P. Borgatti, What’s different about social media networks?: A framework and research agenda, MIS Quart. 38 (2014) 275–304. [11] J. Kleinberg, The convergence of social and technological networks, Commun. ACM 51 (2008) 66–72. [12] C. Haythornthwaite, Social network analysis: an approach and technique for the study of information exchange, Libr. Inform. Sci. Res. 18 (1996) 323–342.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx [13] K. Berger, J. Klier, M. Klier, A. Richter, ‘‘Who is key. . .?’’ – Value adding users in enterprise social networks, in: Proceedings of the 22nd European Conference on Information Systems (ECIS), Tel Aviv, 2014. [14] M. Chau, J. Xu, Business intelligence in blogs: understanding consumer interactions and communities, MIS Quart. 36 (2012) 1189–1216. [15] K. Lerman, R. Ghosh, Information contagion: an empirical study of the spread of news on Digg and Twitter social networks, in: Fourth International AAAI Conference on Weblogs and Social Media, Washington, DC, USA, 2010, pp. 90–97. [16] R.F. Ciriello, D. Hu, G. Schwabe, Identifying patterns of idea diffusion in innovator networks, in: Proceedings of the International Conference on Information Systems, Milan, Association for Information Systems, 2013. [17] O. Oh, M. Agrawal, R. Rao, Community intelligence and social media services: a rumor theoretic analysis of tweets during social crises, MIS Quart. 37 (2013) 407–426. [18] R. Hillmann, M. Trier, Dissemination patterns and associated network effects of sentiments in social networks, in: 2012 IEEE/ ACM International Conference on Advances in Social Networks Analysis and Mining, 2012. [19] J. Moody, D. McFarland, S. Bender-deMoll, Dynamic network visualization, Am. J. Sociol. 110 (2005) 1206–1241. [20] X. Zhang, V. Venkatesh, Explaining employee job performance: the role of online and offline workplace communication networks, MIS Quart. 37 (2013) 695–722. [21] L. Wu, Social network effects on productivity and job security: evidence from the adoption of a social networking tool, Inform. Syst. Res. 24 (2013) 30–51. [22] Y. Wang, M. Iliofotou, M. Faloutsos, B. Wu, Analyzing communication interaction networks (CINs) in enterprises and inferring hierarchies, Comput. Netw. 57 (2013) 2147–2158. [23] X. Zhang, C. Wang, Network positions and contributions to online public goods: the case of Chinese Wikipedia, J. Manage. Inform. Syst. 29 (2012) 11–40. [24] S. Sarker, M. Ajuja, S. Sarker, S. Kirkeby, The role of communication and trust in global virtual teams: a social network perspective, J. Manage. Inform. Syst. 28 (2011) 273–309. [25] M. Trier, Towards dynamic visualization for understanding evolution of digital communication networks, Inform. Syst. Res. 19 (2008) 335–350. [26] J. Howison, A. Wiggins, K. Crowston, Validity issues in the use of social network analysis with digital trace data, J. Assoc. Inform. Syst. 12 (2011) 767–797. [27] J.C. Greene, V.J. Caracelli, W.F. Graham, Toward a conceptual framework for mixed-method evaluation designs, Educ. Eval. Policy Anal. 11 (1989) 255–274. [28] A. Bryman, Integrating quantitative and qualitative research: how is it done?, Qual Res. 6 (2006) 97–113. [29] P.J. Ågerfalk, Embracing diversity through mixed methods research, Eur. J. Inform. Syst. 22 (2013) 251–256. [30] J. Mingers, The paucity of multimethod research: a review of the information systems literature, Inform. Syst. J. 13 (2003) 233–249. [31] S. O’Riordan, J. Feller, T. Nagle, Exploring the affordances of social network sites: an analysis of three networks, in: ECIS 2012 Proceedings, 2013. [32] A. Majchrzak, C. Wagner, D. Yates, The impact of shaping on knowledge reuse for organizational improvement with wikis, MIS Quart. 37 (2013). 455–A12. [33] J.W. Creswell, V. Plano Clark, Designing and Conducting Mixed Methods Research, second ed., Sage Publications, Los Angeles, 2011. [34] V. Venkatesh, S.A. Brown, H. Bala, Bridging the qualitative–quantitative divide: guidelines for conducting mixed methods research in information systems, Manage. Inform. Syst. Quart. 37 (2013) 21–54. [35] N.K. Denzin, The Research Act in Sociology: A Theoretical Introduction to Sociological Methods, (Publishers) Limited, Butterworth, 1970. [36] K.M. Eisenhardt, Building theories from case study research, Acad. Manage. Rev. 14 (1989) 532–550. [37] Association for Information Systems, Senior Scholars’ Basket of Journals. (accessed 10.01.14). [38] J. Vom Brocke, A. Simons, B. Niehaves, K. Riemer, R. Plattfaut, A. Cleven, Reconstructing the giant: on the importance of rigour in documenting the literature search process, in: 17th European Conference on Information Systems, Hampton Press, 2009, pp. 1–13. [39] Z. Shi, H. Rui, A.B. Whinston, Content sharing in a social broadcasting environment: evidence from Twitter, MIS Quart. 38 (2014). 123-A6-A6. [40] N.E. Friedkin, Information flow through strong and weak ties in intraorganizational social networks, Soc. Netw. 3 (1982) 273–285.

17

[41] M. Granovetter, The strength of weak ties: a network theory revisited, Sociol. Theory 1 (1983) 201–233. [42] R. Garg, M.D. Smith, R. Telang, Measuring information diffusion in an online community, J. Manage. Inform. Syst. 28 (2011) 11–38. [43] S. Rios, F. Aguilera, F. Bustos, T. Omitola, N. Shadbolt, Leveraging social network analysis with topic models and the semantic web, in: 2011 IEEEWICACM International Conferences on Web Intelligence and Intelligent Agent Technology, vol. 3, 2011, pp. 339–342. [44] E. Quintane, P.E. Pattison, G.L. Robins, J.M. Mol, Short- and long-term stability in organizational networks: temporal structures of project teams, Soc. Netw. 35 (2013) 528–540. [45] S. Stieglitz, L. Dang-Xuan, Emotions and information diffusion in social media-sentiment of microblogs and sharing behavior, J. Manage. Inform. Syst. 29 (2013) 217–248. [46] Á. García-Crespo, R. Colomo-Palacios, J.M. Gómez-Berbís, B. RuizMezcua, SEMO: a framework for customer social networks analysis based on semantics, J. Inform. Technol. 25 (2010) 178–188. [47] R.L. Cross, A. Parker, The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations, Harvard Business Press, 2004. [48] P.H. Gray, S. Parise, B. Iyer, Innovation impacts of using social bookmarking systems, MIS Quart. 35 (2011) 629–643. [49] McKinsey Global Institute, The Social Economy: Unlocking Value and Productivity through Social Technologies, 2012. [50] G.C. Kane, S.P. Borgatti, Centrality-IS proficiency alignment and workgroup performance, MIS Quart. 35 (2011) 1063–1078. [51] J. Scott, Social Network Analysis, third ed., SAGE, London, 2013. [52] I. Guy, U. Avraham, D. Carmel, S. Ur, M. Jacovi, I. Ronen, Mining expertise and interests from social media, in: Proceedings of the 22nd International Conference on World Wide Web, Geneva, 2013, pp. 515–526. [53] C. Steinfield, J.M. Dimicco, N.B. Ellison, C. Lampe, Bowling online: social networking and social capital within the organization, in: Proceedings of the Fourth International Conference on Communities and Technologies, ACM, New York, NY, USA, 2009, pp. 245–254. [54] R.D. McPhee, P. Zaug, The communicative constitution of organizations: a framework for explanation, Electron. J. Commun. 10 (2000) 1–16. [55] S. Blaschke, D. Schoeneborn, D. Seidl, Organizations as networks of communication episodes: turning the network perspective inside out, Org. Stud. 33 (2012) 879–906. [56] B. Corominas-Murtra, J. Goñi, R.V. Solé, C. Rodríguez-Caso, On the origins of hierarchy in complex networks, Proc. Natl. Acad. Sci. USA 110 (2013) 13316–13321. [57] E. West, D.N. Barron, J. Dowsett, J.N. Newton, Hierarchies and cliques in the social networks of health care professionals: implications for the design of dissemination strategies, Soc. Sci. Med. 48 (1999) 633– 646. [58] L. Page, S. Brin, R. Motwani, T. Winograd, The PageRank Citation Ranking: Bringing Order to the Web, 1998. [59] Z. Shi, A.B. Whinston, Network structure and observational learning: evidence from a location-based social network, J. Manage. Inform. Syst. 30 (2013) 185–212. [60] H. Chen, Chiang, H.L. Roger, V.C. Storey, Business intelligence and analytics: from big data to big impact, MIS Quart. 36 (2012) 1165– 1188. [61] L. Kumar, H. Singh, R. Kaur, Web analytics and metrics: a survey, in: Proceedings of the International Conference on Advances in Computing, Communications and Informatics, ACM, New York, NY, USA, 2012, pp. 966–971. [62] X. Wang, S. Lantzy, A systematic examination of member turnover and online community health, in: Proceedings of the International Conference on Information Systems, ICIS 2011, Shanghai, Association for Information Systems, 2011. [63] N. Poggi, V. Muthusamy, D. Carrera, R. Khalaf, Business process mining from e-commerce web logs, in: F. Daniel, J. Wang, B. Weber (Eds.), Business Process Management, Springer, Berlin Heidelberg, 2013, pp. 65–80. [64] K. Krippendorff, Content Analysis: An Introduction to Its Methodology, third ed., SAGE, Los Angeles, London, 2013. [65] Lasswell, Harold Dwight, Power, and Personality, Norton, New York, 1948. [66] B. Pang, L. Lee, C.C. Aggarwal, C. Zhai, Opinion mining and sentiment analysis, Found. Trends Inform. Ret. 2 (2008) 1–135. [67] M. Ben-Dov, R. Feldman, Text mining and information extraction, in: O. Maimon, L. Rokach (Eds.), Data Mining and Knowledge Discovery Handbook, Springer-Verlag, New York, 2005, pp. 801–831. [68] J. Yates, W.J. Orlikowski, K. Okamura, Explicit and implicit structuring of genres in electronic communication: reinforcement and change of social interaction, Org. Sci. 10 (1999) 83–103.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025

18

S. Behrendt et al. / Computer Networks xxx (2014) xxx–xxx

[69] K. Riemer, A. Richter, Social software: agents for change or platforms for social reproduction? A case study on enterprise microblogging, in: Proceedings of the 21th Australasian Conference on Information Systems, Brisbane, 2010. [70] M. Schnegg, Die Wurzeln der Netzwerkforschung, in: C. Stegbauer, R. Häußling (Eds.), Handbuch Netzwerkforschung, VS Verlag für Sozialwissenschaften, Wiesbaden, 2010. [71] U. Schultze, M. Avital, Designing interviews to generate rich data for information systems research, Inform. Org. 21 (2011) 1–16. [72] S. Kvale, Doing Interviews, Sage Publications, London, 2007. [73] R.K. Yin, Case Study Research: Design and Methods, Sage Publications, 2003. [74] C.-Y. Lin, L. Wu, Z. Wen, H. Tong, V. Griffiths-Fisher, L. Shi, D. Lubensky, Social Network Analysis in Enterprise, Proc. IEEE 100 (2012) 2759–2776. [75] C. Bülow, J. Metscher, J.-M. Schnurr, M. Koch, Shaping a social software for a distributed military organisation, in: Proceedings of the 13th International Conference on Knowledge Management and Knowledge Technologies, Granz, Austria, 2013. [76] A. Richter, K. Riemer, Malleable end-user software, Bus. Inform. Syst. Eng. 5 (2013) 195–197. [77] NATO, NATO Codes for Grades of Military Personnel: STANAG 2116, NATO, 1996. [78] R.S. Burt, Structural Holes: The Social Structure of Competition, Harvard University, Cambridge (Mass), London, 1992. [79] W. Fan, L. Wallace, S. Rich, Z. Zhang, Tapping the power of text mining, Commun. ACM 49 (2006) 76–82.

Sebastian Behrendt is a Ph.D. candidate at Cooperation Systems Center at Bundeswehr University Munich. He graduated in Information Management at the University of Applied Science and Arts Hannover in 2006. After that, he worked several years as a consultant and project manager in the internet industry before going back to the academics. Now he is focusing on the evaluation of social software in organizations. In a current research project he analyzes the development of informal networks in such tools and their influence on knowledge exchange. He furthermore co-developed a framework that has supported the selection, implementation and evaluation of social software at different companies like Allianz or Bosch.

Alexander Richter is senior researcher at the Information Management Research Group at University of Zurich and head of Social Business at Cooperation Systems Center at Bundeswehr University Munich. In the last years he has been working in engaged scholarship projects on national and European level to understand the role of social software in the increasingly differentiated working practices of knowledge workers. By doing so, he studies the impact of social platforms in emergent structures such as informal networks, and new forms of leadership and innovation. Furthermore, he is interested in the development of methods to evaluate these developments and approaches to address them. He has for example co-developed a framework that has supported the selection, implementation and evaluation of social software at companies such as Allianz, Airbus, Bayer and Bosch. He is Chairman of the CSCW SIG in the German computer society and a mentor of a startup in the domain of ambient assisted living.

Matthias Trier is associate professor at Copenhagen Business School. He studies phenomena related to electronic communication and social influence effects in online media within and outside the organization with a mixed methods approach that blends quantitative, qualitative and network analytical methods. Example topics include the implementation/appropriation of social media, online participation, framing electronic discourses (e.g. from a management perspective), information transfer, dissemination processes or bottom-up community emergence as a part of knowledge management initiatives. One special methodological focus is on developing an event-driven method for dynamic network analysis. The corresponding software Commetrix enables research into emerging structures and dynamic processes of networking among people.

Please cite this article in press as: S. Behrendt et al., Mixed methods analysis of enterprise social networks, Comput. Netw. (2014), http:// dx.doi.org/10.1016/j.comnet.2014.08.025